AI Recurring Payment Failure Predictor

Ilvix Works

Remote · $100k – $130k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

A subscription business that runs a monthly charge and gets declined has a problem it can handle in several ways: retry immediately, retry at a different time of day, retry with a different retry interval, pre-emptively reach out to the customer, or update the payment method. Which approach works best depends on why the charge failed — insufficient funds, expired card, bank-side fraud hold, or network error — and predicting that reason before the retry is what smart dunning management looks like. We process recurring payments for 85 subscription businesses with combined annual recurring revenue of $4B. Payment failure prediction is the ML capability that makes our dunning management intelligent. We can predict payment failure probability 3-5 days before a charge attempt with 78% precision. We can't yet predict failure reason — which is what drives the dunning strategy selection. You'll build the failure reason prediction system: engineering features from prior payment history, bank response codes, account behavior, and macro signals (end-of-month balance pressure), training multi-class failure reason models, and integrating predictions into the dunning strategy selection logic that decides retry timing, channel, and intervention type for each subscriber.

Fully remote.

Responsibilities

Build multi-class payment failure reason prediction: insufficient funds, expired card, fraud hold, network error
Engineer features from payment history, bank response patterns, subscriber behavior, and calendar signals
Integrate failure reason predictions into dunning strategy selection for retry timing and channel decisions
Develop prediction calibration ensuring probability estimates are reliable for strategy thresholds
Create monitoring tracking failure reason prediction accuracy against realized bank response codes
Build failure reason trend analysis identifying systematic shifts in decline composition

Requirements

3+ years in data science or ML with payments, subscription analytics, or financial services experience
Strong Python, SQL, and scikit-learn skills for structured data classification
Understanding of payment processing, bank response codes, and subscription billing dynamics
Experience building classification systems with calibrated probability outputs
Ability to integrate ML predictions into operational workflow decision logic